C-type lectin-like domains

Abstract

Carbohydrate-recognition domains of C-type (Ca²⁺-dependent) animal lectins serve as prototypes for an important family of protein modules. Only some domains in this family bind Ca²⁺ or sugars. A comparison of recent structures of C-type lectin-like domains reveals diversity in the modular fold, particularly in the region associated with Ca²⁺ and sugar binding. Some of this diversity reflects the changes that occur during normal physiological functioning of the domains. C-type lectin-like domains associate with each other through several different surfaces to form dimers and trimers, from which ligand-binding sites project in a variety of different orientations.

Introduction

C-type animal lectins represent an important recognition mechanism for oligosaccharides at cell surfaces, attached to circulating proteins and in the extracellular matrix. Binding of specific sugar structures by these lectins mediates biological events, such as cell–cell adhesion, serum glycoprotein turnover and innate immune responses to potential pathogens [1]. Proteins with diverse overall architecture contain homologous carbohydrate-recognition domains (CRDs) that mediate sugar binding. A comparison of the sequences of these modules defines a common sequence motif, whereas structural analysis of a prototypical domain reveals that this motif largely reflects the importance of the conserved residues in establishing the fold of this type of module [2].

Evidence discussed here and elsewhere indicates that many protein modules containing part or all of the C-type CRD motif serve functions other than saccharide recognition. Hence, it is appropriate to consider this motif a characteristic of C-type lectin-like domains (CTLDs) to reflect their similarity to the CRDs of C-type lectins without necessarily implying common function [3]. The structures of roughly a dozen CTLDs have now been established. Although overall similarity in fold is the most obvious feature of these structures, comparisons also reveal important differences in fold, in interactions with Ca²⁺ and in oligomerisation. The results summarised in this review highlight some of the patterns that have emerged, using the previously described structure of the CRD from rat serum mannose-binding protein as a prototype for the superfamily.

Convergent and divergent evolution the C-type lectin-like domain fold

Figure 1 summarises two important features of CTLDs that have emerged from recent structural work. First, a topology similar to the fold originally identified in mannose-binding protein has been recognised in at least three types of module that are not evidently related at the sequence level: the link-protein-type module from the mammalian cell-surface molecule CD44 [4], the angiogenesis inhibitor endostatin [5] and the bacterial adhesion molecule intimin [6]. The lack of sequence similarity and important differences in the packing of the hydrophobic cores of these molecules are strong indicators that the topological similarity results from convergent evolution. Second, extensive structural divergence within the CTLD superfamily is also documented, reflecting a range of functions that has become evident.

Amongst the structures of CTLDs that have been described in the past three years, most of the elements of regular secondary structure are conserved, although there is considerable variation in the loop structures. The most striking departure from the prototype mannose-binding protein structure is the absence of one α helix in the CTLD of the natural killer cell receptor CD94 [7^••]. The sequence of this segment retains the pattern of hydrophobic amino acids that would be expected to generate an amphipathic helix suitable to form the side of the domain and sequence alignments would have led to the prediction that the helix would be present in CD94. As discussed below, its absence apparently reflects alternative interactions in the dimer interface formed by this portion of the domain.

Earlier structures of CRDs from mannose-binding protein and E-selectin were both of relatively short CRDs lacking an N-terminal disulfide bond present in many CTLDs 8, 9, 10. An extra segment appears in several more recently determined structures of CTLDs that are sometimes referred to as long-form CRDs or CTLDs. These longer CTLDs include lithostathine [11], tetranectin [12], factors IX/X-binding protein [13^••] and CD94 [7^••]. As expected from the presence of a disulfide bond linking cysteines separated by 10 residues in this segment, it forms a hairpin in all the structures. In several cases, one portion of the hairpin takes on β structure and pairs with the most N-terminal conserved β strand seen in all CTLDs.

The recently completed Caenorhabditis elegans genome sequence [14] provides an opportunity to explore the extent of the CTLD radiation. Roughly 180 potential CTLDs have been identified [15]. Of these, fewer than 10% are predicted to bind Ca²⁺ or carbohydrates, based on the features essential for such functions in mammalian CTLDs. Interestingly, the domain architectures of most C. elegans proteins containing CTLDs also differ from those of the known structural classes of mammalian proteins containing CTLDs. This difference may reflect the fact that important classes of mammalian proteins containing CTLDs remain to be identified, but it also may be a result of extensive shuffling of protein modules that has taken place since the divergence of the vertebrate and invertebrate lineages.